UNIVERSIDADE FEDERAL DE SÃO CARLOS
CENTRO DE CIÊNCIAS EXATAS E DE TECNOLOGIA
PROGRAMA DE PÓS-GRADUAÇÃO EM ENGENHARIA DE
PRODUÇÃO
GUSTAVO BAGNI
CARD-BASED SYSTEMS: SYSTEMATIC LITERATURE REVIEW
OF NEW SYSTEMS AND PROPOSAL OF A LIST OF SOFT
FACTORS FOR SYSTEMS IMPLEMENTATION
São Carlos-SP
2019
GUSTAVO BAGNI
CARD-BASED SYSTEMS: SYSTEMATIC LITERATURE REVIEW OF NEW SYSTEMS AND PROPOSAL OF A LIST OF SOFT FACTORS FOR SYSTEMS IMPLEMENTATION
Dissertation presented to the Graduate Program in Production Engineering of the Federal University of São Carlos, to obtain the title of Master in Production Engineering. Supervisor: Prof. PhD Moacir Godinho Filho
São Carlos-SP 2019
DEDICATION
To my beloved parents, Claudinéia and Orlando, and brother, Guilherme, who gave me strength and continually provided their support along this journey.
ACKNOWLEDGMENTS
To my parents, Orlando and Claudinéia, who supported me unconditionally in this challenge as in many others, providing all the support I needed.
To my brother Guilherme who inspired me to go back to university to continue studying. To my grandparents, Renato (in memoriam) and Dalva, Orlando and Teresinha, for their wisdom to provide better conditions to their descendants.
To my advisor, professor Moacir, for his guidance in this research and valuable initial suggestion to write this dissertation in English and in articles format.
To professors Matthias Thürer, Mark Stevenson and Gilberto Ganga (Giba), for the contributions and suggestions that have significantly improved this research.
To my friend Josadak, for teaching me how to write my first articles.
To all UFSCar professor who have contributed to my education since graduation.
To my master’s degree colleagues, for the numerous conversations and discussions that helped to clarify my research theme.
To my colleagues and friends at work, who made it possible for me to leave the company to attend classes regardless of the problem we were facing.
RESUMO
Embora os Sistemas de Coordenação de Ordens sejam considerado um tópico maduro na literatura, o foco da literatura até o momento recaiu especialmente sobre a otimização dos parâmetros de funcionamento de cada sistema, atribuindo pouca importância aos fatores relacionados a influência humana (soft factors). A implementação, contudo, continua a ser um problema complexo. Esse trabalho visa reduzir essa lacuna, propondo, através de uma combinação de métodos de pesquisa, uma lista de fatores soft críticos para o sucesso na implementação de sistemas baseado em cartão, os quais são os mais estudados e implementados e que compartilham como característica uma forte influência humana em seu funcionamento. Contudo, para realização desse objetivo, inicialmente foi necessário identificar quais são os sistemas baseados em cartão. Embora para sistemas como o Kanban e o CONWIP exista uma grande literatura disponível, sistemas desenvolvidos após a proposta do POLCA em 1998 foram pouco explorados. Assim, essa dissertação inicialmente realiza uma revisão sistemática de literatura identificando 13 sistemas desenvolvidos entre 1999 e 2018, tais como COBACABANA, DDMRP, Redutex, B-CONWIP, BK-CONWIP, dentre outros. Brevemente, é apresentado o funcionamento, características, estágio atual de pesquisa e ambientes propícios para cada sistema, visando aumentar as pesquisas sobre eles. Os sistemas são também comparados em relação a seis variáveis, identificando-se que muitos dos novos sistemas são baseados em cartão. Os fatores soft propostos para a implementação de sistemas baseado em cartão se baseiam na análise de problemas citados na literatura bem como de dificuldades identificadas através de um estudo de caso longitudinal. Essa lista foi validada por especialistas assim como por um grupo de colaboradores da empresa foco que participou da implementação do kanban. Nessa dissertação é proposta também uma casa de fatores soft para a implementação de sistemas baseados em cartão, nas quais os fatores são classificados como exclusivos dessa temática ou fatores clássicos de administração, bem como em relação ao nível organizacional em que atua. (organização, grupo de implementação ou indivíduo). Essa casa tem como objetivo auxiliar os gerentes na implementação de sistemas baseados em cartão, aumentando as taxas de sucesso nesse processo. Além disso, através da revisão de sistemas de coordenação de ordens recentes, essa dissertação visa aumentar o repertório dos gerentes sobre os sistemas existentes, possibilitando a implementação de opções mais adequadas para o ambiente produtivo em que se encontram.
Palavras-chave: Sistemas de coordenação de ordens. Sistemas baseado em cartão. Fatores soft.
ABSTRACT
Although Production Control Systems are considered a mature topic in literature, up to now, the focus of literature has been on optimizing the operating parameters of each system, assigning little importance to factors related to human influence (soft factors). Implementation, however, remains a complex problem. This paper aims to reduce this gap by proposing, through a combination of research methods, a list of soft factors critical to success in implementing card-based systems, which which are the most studied and implemented and which share as a characteristic a strong human influence on their functioning. However, to achieve this goal, it was initially necessary to identify which systems are based on cards. Although for systems such as Kanban and CONWIP there is a large literature available, systems developed after POLCA proposal in 1998 were little explored. Therefore, this dissertation initially performs a systematic literature review identifying 13 systems developed between 1999 and 2018, such as COBACABANA, DDMRP, Redutex, B-CONWIP, BK-CONWIP, among others. Briefly, it presents how each system works, its characteristics, current research stage and environments in which it has been proved to be useful, aiming to increase researches about them. The systems are also compared in relation to six variables defined in the literature, identifying that many of the new systems are card-based. The soft factors proposed for the implementation of card-based systems are based on the analysis of problems cited in the literature as well as difficulties identified through a longitudinal case study. This list was validated by experts as well as a group of employees from the focus company that participated in the implementation of kanban. This dissertation also proposes a soft factor house for the implementation of card-based systems, in which the factors are classified as exclusive to this theme or classic management factors, as well as in relation to the organizational level in which it operates (organization, implementation group, or individual). This house aims to assist managers in implementing card-based systems, increasing success rates in this process. In addition, by reviewing recent production control systems, this dissertation aims to increase the repertoire of managers on existing systems, enabling the implementation of more appropriate options for the productive environment in which they are located.
Keywords: Production Control System. Card-based system. Soft factors. Kanban. COBACABANA, POLCA. DDMRP
LIST OF FIGURES
Figure 1 – Overview of Dissertation Structure ... 17
Figure 2- Systematic Literature Review A ... 21
Figure 3- Production Control Systems evolution from 1999 to 2018 ... 26
Figure 4 - Base Stock and Inverse Base Stock ... 27
Figure 5 - Customised token-based systems ... 27
Figure 6 - Behaviour-Based Control authorizations ... 29
Figure 7 – Gated MaxWIP ... 29
Figure 8 - Parallel Pull Flow... 30
Figure 9 - COBACABANA card loop ... 31
Figure 10 - COBACABANA planning board ... 32
Figure 11 - BK-CONWIP ... 34 Figure 12 - B-CONWIP ... 35 Figure 13 - DSSPL ... 36 Figure 14 - DDMRP ... 37 Figure 15 – Redutex ... 39 Figure 16 – CONLOAD ... 40 Figure 17 – DEWIP ... 41
Figure 18 - Combination of research methods used in this study ... 49
Figure 19 – Research String ... 52
Figure 20 - Systematic Literature Review B ... 52
LIST OF TABLES
Table 1 - Research Protocol A ... 20
Table 2 - The 13 PCS’s classified according to the four variables of system’s characteristics dimension ... 25
Table 3 - Research gaps and future research directions ... 44
Table 4 - Research Protocol B ... 51
Table 5 - Systematic Literature Review Results ... 53
Table 6 - Case study: interviewees’ characteristics ... 56
Table 7 - Case study: experts’ characteristics ... 59
Table 8 - Experts Panel: procedure to revise the initial proposal ... 60
Table 9 – Soft factors list refined by experts ... 67
Table 10 – Comparison between card-based systems and Lean soft factors... 69
LIST OF ABBREVIATIONS
BBC Behaviour Based Control
B-CONWIP Basestock-Constant Work-in-Process
BK-CONWIP Basestock Kanban-Constant Work-in- Process COBACABANA Control of Balance by Card Based Navigation CONLOAD CONstant LOAD
CONWIP Constant Work-In-Process CSF Critical Success Factors
CTBS Customize Token-based system
DBR Drum-Buffer-Rope
DDMRP Demand Driven Materials Requirement Planning DEWIP Decentralized Work in Process
D-KDP Dedicated Kanban Distribution Policy DSSPL Double Speed Single Production Line G-MaxWIP Gated MaxWIP
HK-CONWIP Hybrid Kanban-CONWIP IBS Inverse Base Stock LCS Longitudinal Case Study MRP Master Requirements Planning PBC Periodic Batch Control
PCS Production Control Systems
POLCA Paired Cell Overlapping Loops of Cards PPC Production Planning and Control
PPF Parallel Pull Flow
RRC Restrictive Resource Capacity S-KDP Shared Kanban Distribution Policy SLR Systematic Literature Review SME Small and Medium Enterprises
SWIP Self-regulated Work-In-Process
WIP Work in Process
TABLE OF CONTENTS
1 INTRODUCTION ... 14
1.1 Contextualization ... 14
1.2 Research Question and Objectives ... 15
1.3 Overview of Research Method ... 16
1.4 Overview of Dissertation Structure ... 16
2 SYSTEMATIC REVIEW AND DISCUSSION OF PRODUCTION CONTROL SYSTEMS THAT EMERGED BETWEEN 1999 AND 2018 ... 18
2.1 Introduction ... 18
2.2 Research Method ... 19
2.1.1 Article Selection ... 20
2.1.2 Content Analysis ... 22
2.3 Results: New Production Control Systems (PCS) ... 24
2.4 Comparison of New Production Control Systems (PCS) ... 42
2.5 Conclusions and Research Agenda ... 43
2.5.1 Conclusions ... 43
2.5.2 Research Agenda ... 44
3 SOFT FACTORS FOR CARD-BASED SYSTEMS IMPLEMENTATION: A MULTI-METHOD STUDY ... 47
3.1 Introduction ... 47
3.2 Method ... 49
3.2.1 An overview of the research method ... 49
3.2.2 Systematic Literature Review (SLR) ... 50
3.2.3 Longitudinal Case Study (LCS) ... 53
3.2.4 Content Analysis ... 57
3.2.5 Experts Panel ... 58
3.2.6 Validation with Company’s Employees ... 60
3.2.7 Validation with Company’s Employees ... 61
3.2.8 Research Quality ... 61
3.3 Results ... 62
3.3.1 An initial list of soft factors based on SLR and LCS ... 62
3.3.2 Refining the list by means of expert panel ... 66
3.4 Discussion ... 68
3.4.1 Findings ... 68
3.4.2 Propositions ... 70
3.4.3 Card-based soft factors house ... 72
3.5 Conclusions and Research Agenda ... 73
3.5.2 Limitations and Research Agenda ... 73
4 CONCLUSIONS ... 75
REFERENCES ... 77
APPENDIX A – CASE STUDY RESEARCH PROTOCOL ... 87
1 INTRODUCTION
1.1 Contextualization
As a significant part of the capital of industrial organizations is in manufacturing, managing these resources efficiently is essential for building or maintaining competitive positions. Therefore, Production Planning and Control (PPC) plays a fundamental role in deploying organization’s strategic plans in manufacturing tactical and operational plans, as well as in connecting production and purchase of materials to customer needs. In the heart of PPC, there are the Production Control Systems (PCS’s), which regulate information and materials flows through the factory (KARRER; ALICKE; GÜNTHER, 2012).
The literature about PCS’s has focus mainly on mathematical approaches to optimize the parameters of each system (PONS; 2010; HENDRY; HANGANG, STEVENSON, 2013). Implementation, however, remains a complex problem (RAZMI; AHMED, 2003). Most implementation studies only describe the system logic inside a business environment, but do not systemically address the difficulties during the implementing phase. For systems based on cards (card-based system), which are the most studied and implanted PCS’s, those difficulties are even more significant as a characteristic shared by those systems is the strong human influence on its operation (SALEM et al., 2006; LIU; HUANG; 2009).
Those difficulties can be related to factors critical to card-based systems implementation. Those factors are referred in literature by different denominations, such as implementation factors (PARZINGER; NATH, 1998; CHOU; CHANG, 2008), project success factors (DVIR et al., 1998), critical success factors (ROCKART, 1979; HOWELL, 2009; NETLAND; 2016), among others. In this dissertation, as the emphasis is on the human influence on card-based systems implementation, these factors will be called soft factors (ABDULLAH; ULI; TARÍ, 2008). Examples of soft factors are workers motivation, support from top management and communication. Therefore, hard factors, such as availability of financial resources, adequacy of equipment, technology, among others, are out of the scope of this study.
Apart from Pons (2010), soft factors have been rarely studied in the specific context of card-based systems (MARODIN; SAURIN, 2013). Therefore, understanding which soft factors are critical to card-based system implementation are key to increase the success on implementing those systems. Moreover, there is no list of soft factors which managers should focus their attention on while implementing a card-based system (ROCKART, 1979; HOWELL, 2009). Therefore, the following research questions arises: Which are the soft factors
critical to card-based systems implementation?
Aiming to reduce this gap, this dissertation provides a list of soft-factors for card-based systems implementation, based on empirical papers and on a longitudinal case study. The final list contains factors as diverse as management support, implementation during low demand period and card’s material quality. In the list, we tried to maintain the factors as generic as possible, not including specific factors suitable for particular environments. Therefore, additional factors can be included for each individual implementation.
Therefore, in terms of research, we seek to highlight the importance of human factors on card-based system implementation, asking for more studies in this field. In terms of practice, we hope that the list helps companies to increase the success in implementing card-based systems and that managers know in advance which soft factors they should concentrated their attention during the implementation process.
However, while doing this research, we identified that it was not clear in literature which are the PCS’ based on cards. Classical systems, such as kanban and CONWIP, were certainly part of this PCS’s group, but what about more recently systems, proposed after 1998, the year that POLCA emerged? Therefore, aiming to identified card-based systems proposed in the last 20 years (from 1999 and 2018), we decided to perform first a SLR in which 13 PCS’s were identified (based on cards or not) (Chapter 2). We briefly present how each system works, its characteristics and environments it is suitable for. To the best of our knowledge, there is no study that summarizes the main advances concerning PCS between 1999 and 2018.
All new PCS’s identified in the SLR were included, although there is a great difference among them regarding evolutionary stage and number of articles published. However, they all can provide interesting insights to the proposal of new PCS’s as well as they can provided elements to understand why some PCS’s have more success than others do. Therefore, this dissertation also contributes to research, outlining new search directions for future research on PCS’s and contributing to disclosure PCS’s proposed recently, and for practice, helping managers to find new solution to their day-to-day problems.
1.2 Research Question and Objectives
From this research question, the following objectives were defined:
1. To identify the Production Control Systems developed over the last 20 years (from 1999 to 2018), their characteristics and environments in which they have been proved to be useful;
3. To identify difficulties in empirical card-based systems implementation related to human aspects;
4. Based on those difficulties, to propose a list of soft factors critical to card-based systems;
5. To refine this list by means of expert panel.
1.3 Overview of Research Method
This dissertation uses a variety of methods to reach the five research objectives: • Systematic Literature Review: used in two different moments. First, to identify
PCS developed between 1999 and 2008. Secondly, to identify difficulties related to card-based systems in empirical papers;
• Case Study: to identify the problems faced by an organization which failed to implement kanban in its final process by a longitudinal study;
• Content Analysis: to analyze all the information difficulties identified in SLR and LCS in order to propose soft factors for card-based systems implementation; • Expert panel: to refine the previous list by 6 experts.
1.4 Overview of Dissertation Structure
This dissertation is structure in four chapters (Figure 1). Chapter 2 and 3 are written in article format, aiming to increase the visibility of research’s results in international journals. Therefore, the author apologizes for eventual redundancies among the chapters.
Chapter 1 briefly presented the context and motivation of this research, as well as research question and objectives. An overview of research method and structure are also provided.
Chapter 2 aims to achieve objectives 1 and 2 of this dissertation. Therefore, a Systematic Literature Review (SLR) were conducted to identify new PCS’s developed between 1998 and 2008, excluding simple extensions of classical PCS, such as variations of Kanban and POLCA. The SLR results in 13 new systems, which were presented in detail regarding how they work, their main characteristics and in which environments they have been proved to be useful. We also analyzed which systems have been further studied after its initial proposal and which ones have been reported to be implemented in real production systems. In addition, we identify that from the 13 PCS, 8 are card-based systems and 3 used cards partially. Therefore, it was defined to focus implementation difficulties of PCS in card-based systems (Chapter 3).
Figure 1 – Overview of Dissertation Structure
Chapter 3 aims to achieve objectives 3, 4 and 5. Therefore, first a SLR were conducted to identify human difficulties related to card-based systems implementation. However, as little difficulties were identified in literature, a longitudinal case study were also conducted in an organization which failed to implement kanban. Performing a content analyzed with the difficulties identified and based on soft factors for lean implementation, a list of soft factors for card-based systems is proposed. Lean soft factors were used as references because there are many studies that proposed list of those factors. Moreover, many card-based systems are based at least partially on Lean approach. Finally, this list was refined by 6 experts which came from university (professor), manufacturers (managers) and consulting (consultants).
Chapter 4 highlights the main contributions of this dissertation, research limitations and proposal for future studies.
1. Introduction
2. PCS developed
between 1999 and 2018
(objetives 1 and 2)
3. Soft Factors for
card-based system
implementation
(objetives 3, 4 and 5)
2 SYSTEMATIC REVIEW AND DISCUSSION OF PRODUCTION CONTROL SYSTEMS THAT EMERGED BETWEEN 1999 AND 2018
Although there is a large literature about classical Production Control Systems (PCS) such as Kanban, CONWIP and Materials Requirements Planning (MRP), few articles deal with recent PCS, specifically the ones proposed after the emerged of the POLCA in 1998. Therefore, in this chapter, a Systematic Literature Review (SLR) is conducted to identified PCS’s proposed over the last 20 years, as well as their characteristics and environments they proved to be suitable for. Moreover, the 13 PCS’s identified are compared regarding 6 variables, in order to identify similarities and differences among them.
2.1 Introduction
Production Control Systems (PCS) are a key factor for effective manufacturing systems as they regulate the information and materials flows through the factory (MASIN; HERER; DAR-EL, 2005; KARRER; ALICKE; GÜNTHER, 2012). Therefore, the choice of an appropriate PCS is an important success factor for any organization (HASSAN; KAJIWARA, 2013). Consequently, many different PCS’s emerged. This includes Kanban systems (e.g. Sugimori et al. (1977), Berkley (1992), Monden (1998) and Lage Junior and Godinho Filho (2010)); Constant Work-In-Process (CONWIP; e.g. Spearman, Woodruff and Hopp (1990), Framinan, Gonzales and Ruiz-Usano (2003), Prakash and Chin (2014) and Jaeglar et al. (2017)); Drum-Buffer-Rope (DBR; e.g. Goldratt (1990), Guide (1996) and Mabin and Balderstone (2003)); Periodic Batch Control (PBC; e.g. Burbidge (1996), Benders and Riezebos (2002)); Materials Requirements Planning (MRP; e.g. Orlicky (1975) and Mohebbi, Choobineh, and Pattanayak (2007)); Workload Control (WLC; e.g. Land and Gaalman (1998) and Land (2006)) and Paired Cell Overlapping Loops of Cards with Authorization (POLCA; e.g. Suri (1998) and Riezebos (2010)).
For these ‘classical’ PCS, there is a large literature available. Some papers present a literature review of different PCS, for example Stevenson, Hendry and Kingsman 2005, Liu and Huang (2009), Fernandes and Godinho Filho (2011) and Thürer, Stevenson and Protzman (2017). There is also a large number of works comparing different PCS by use of simulation, for example, Liu and Huang (2009), Koulouriotis, Xanthopoulos and Tourassis (2010), Sato and Khojasteh-Ghamari (2012), Silva et al. (2017) and Thürer et al. (2019).
However, all of this literature focusses on PCS’s developed before 1998, the year when POLCA emerged. To the best of our knowledge, there is no study that summarizes the main
advances concerning PCS in the last 20 years. Thürer, Stevenson and Protzman (2017), for example, only included card-based systems on their review. In response to this gap, this study provides a review and discussion of new PCS’s that emerged since 1998. Our definition of new PCS’s excludes extensions of classical PCS (for example, variations of Kanban, ConWIP and POLCA) that do not significantly change the nature of the original system. Therefore, the systems included in this chapter either differ significantly from existing ones (such as COBACABANA) or combine elements and characteristics of two or more existing systems (such as BK-CONWIP and B-CONWIP).
All new PCS’s identified in the Systematic Literature Review were included in this paper, although there is a great difference among them regarding evolutionary stage and number of articles published. However, they all can provide interesting insights to the proposal of new PCS’s as well as they can provided elements to understand why some PCS’s have more success than others do. In terms of research, we seek to outline new search directions for future research on PCS’s. In terms of practice, we hope that our study helps managers to find new solution to their day-to-day problems.
The remainder of this chapter is organized as follows. Section 2.2 describes the research method used in this chapter, Systematic Literature Review, and the main variables defined to compare different PCS’s. In Section 2.3, the PCS identified in the SLR are described, with emphasis on how they work and the most suitable environments for each of them. Section 2.4 compares the PCS identified in the SLR according to their evolution and the variables defined in Section 2. Finally, Section 2.5 provides some conclusion arguments, limitations and suggestions for future researches.
2.2 Research Method
This study started by asking:
RQ What are the characteristics of Production Control Systems that newly emerged in the last 20 years?
A systematic literature review is considered the most adequate method to answer our question since it allows for understanding existing knowledge in more depth while minimizing bias in the selection of articles (TRANFIELD; DENIER; SMART, 2003; FAWCETT et al., 2014). The two subsections below outline the approach adopted for article selection and analyzes of the articles.
2.1.1 Article Selection
Article selection followed the four steps proposed by Tranfield, Denier and Smart (2003) and used in many articles, like Negrão, Godinho Filho and Marodin (2017), which are: • Step 1 – Search in database: following the protocol presented in Table 1, search was conducted in Web of Science and Scopus. According to Thomé, Hollmann and Scavarda (2014), a SLR should search in, at least, two databases. Web of Science and Scopus were chosen because they are regularly updated and cover a wide breath of subjects (CHADEGANI et al. 2013; THOMÉ; SCAVARDA; SCAVARDA, 2016). The research results in 955 non-duplicated articles.
Table 1 - Research Protocol A
Research Protocol
Database Web of Science and Scopus Publication Years From 1999 to 2018
Document type Journals
Language English
Strings “production control system*” “production system” AND “push*” “production system*” AND “pull” “card based” AND “production” “production system” AND “hybrid” “production control” AND “pull” “production control” AND “push*” Inclusion criteria • Articles featuring a new PCS
• Applications or comparisons of PCS developed over the last 20 years
Exclusion criteria • Evolution of classical systems, such as Kanban and CONWIP;
• Application of sequencing rules to prioritize production; • Review literature of existing PCS
Source: Elaborated by the author.
• Step 2 – First filter: The title and the summary of the 955 articles were evaluated in order to assess whether they met inclusion and exclusion criteria of the research protocol. First of all, as our objective is to identified PCS’s proposed from 1999 to 2018, papers which only review systems developed before 1999 were excluded. Papers which presents evolution of classical systems were also excluded because these systems do not match the definition of new PCS’s presented in the introduction of this chapter. Moreover, PCS’s are much larger than sequencing rules, which optimize work sequence in a specific work station
or group of stations. Therefore, papers, which only deal with sequencing rules, were also excluded. Therefore, our inclusion criteria are papers which present proposals of new PCS (theoretical, simulation or empirical) or that compared PCS’s in which at least one of them were developed from 1999 to 2018. The first filter resulted in 124 articles.
• Step 3 – Second filter: This filter consists of full reading of the 124 remaining articles, again applying inclusion and exclusion criteria defined in the research protocol. The second filter resulted in 24 articles. These three steps are summarized in Figure 2.
• Step 4 – Final Selection: To the 24 articles selected, 12 more were added using the snowball approach, resulting in 36 articles. Some articles were identified by citations in the 24 articles that resulted from the SLR. Others were included by searching for the name of the PCS’s identified in the databases. Snowball approach added a significant number of articles to this review because many papers use specific key words to propose new PCS’s, such as lot release rule, manufacturing control, production line control, materials management, among others. Therefore, we were not able to define a group of keywords that would systematically result in most of the 36 articles. Instead, we selected keywords that would result in the majority of them, and the others were identified by snowball approach.
Figure 2- Systematic Literature Review A
Source: Elaborated by the author.
Search in electronic databases 1stfilter 2ndfilter 995 117 WEB OF SCIENCE SCOPUS 107 211 637 WEB OF SCIENCE SCOPUS 2 15 7 WEB OF SCIENCE SCOPUS 17 54 46 12 Snowball Approach 36 24
2.1.2 Content Analysis
This stage involved extracting and documenting information from the 36 sources. To minimize subjectivity, the author: (i) cross-checked results; and, (ii) conducted regular meetings to resolve any emerging inconsistencies in interpreting the results. From our sample, 13 PCS were identified.
As a template for data collection, a simple matrix was used where, for each PCS (row), we asked (column):
• What are the characteristics of the system? • How does the system work?
• In what productive environments did this system prove useful? • What research was available involving the system?
To compare the 13 PCS’s, two dimensions were selected: systems’ characteristics and evolution. Regarding systems’ characteristics, four variables were selected, as follows:
(1) Primary Control Variable: a system can either control WIP (Work in Process) or throughput. If a PCS control WIP, then it observes throughput. The opposite is also true (HOPP; SPEARMAN, 2008);
(2) Degree of Centralization: if order release is controlled by a central entity (e.g. production planner), than the PCS is Centralized. An example of a classical centralized PCS is MRP, as all orders are release by production planning. Local stations only execute the order. On the other hand, some systems are Decentralized, because the local production stations are responsible for defining when to start an order an even which order to start (not the central planning). This occurs, for example, in Kanban. In some PCS, there are more than one type of release authorizations. For example, in BK-CONWIP, an order is processed only if received ConWIP, Base Stock and Kanban authorization. It is possible that some of these authorizations are centralized and some are decentralized. Therefore, those systems are classified as mixed. For example, in BK-CONWIP, Kanban authorizations are decentralized (locally controlled by production stations), but ConWIP and Base Stock authorizations are centralized (controlled by the central production planning);
(3) Suitability to material flows: it is important to understand to what kind of environment a PCS is more suitable for in order to choose a more adherent system to the environment analysed. An important variable in the shop floor is
the flow of materials. Some PCS are more suitable to flow shop (flow of items occur in the same direction) and others to job shop (flow of items occurs in different direction) (Johnson and Montgomery 1971). However, in some cases, although a PCS is more adequate to a certain type of flow, it can also be applied in the other. An example is COBCABANA, initially proposed for job shops, but in later papers simulated successfully for flow shops;
(4) Card-based system: a PCS is classified as card-based if it was originally introduced based on card signals. But note that these signals can also be other physical entities (such as boxes) or even electronic signals (Thürer, Stevenson, and Protzman 2017). As DSSPL, DSSPL, DDMRP and Redutex were introduced using cards to trigger the work of some items or some production stages, they are classified as partially. As cards are used to control the stock or workload levels of the systems, not the throughput, all card-based systems identified in the review have WIP as primary variables. The opposite, however, is not true (all systems that have WIP as primary variable are not card-based). CONLOAD is an example, as it controls WIP, but does not use cards.
The first and second variables (primary control variable and degree of centralization) are presented by Lödding, Yu and Wiendahl (2003). The third (material flow) is adapted from Löoding, Yu and Wiendahl (2003). Originally, these authors classified the system flow complexity into high and low. However, given the predominance in literature of job shop and flow shop concepts, we will use these classes for the intermittent systems presented, as proposed by Johnson and Montgomery (1971). The two classifications, however, are integrated, since the materials flow of a job shop system is more complex and of a flow shop is simpler. Finally, the fourth variable (card-based systems) was included, given the importance card-based systems received over the last two decades in the literature and its wide application in real systems, especially for its implementation simplicity and visual control (LIBEROPOULOS; DALLERY, 2000; THÜRER; STEVENSON; PROTZMAN, 2017).
Understanding evolution of the systems is interesting because it can provide important insights about patterns to suggest a new PCS. In this dimension, two variables were selected:
(1) Number of articles published about a PCS’s: we considered only papers that contribute clearly to the development of a PCS by a mathematical simulations, empirical application or comparison with other systems. Therefore, papers that only cite the system were not considered in the evolution analysis;
into axiomatic or empirical, following the classification of Operations Management papers proposed by Bertrand and Fransoo (2002). This variable was important to further analyses how closed are literature and practice regarding the new systems identified.
2.3 Results: New Production Control Systems (PCS)
Table 2 presents the 13 PCS’s identified in the SLR as well as classify them regarding the four variables of system’s characteristic’s dimensions. Figure 3, presents the systems evolution, with all 36 papers founded in the SLR about each of the 13 PCS’s as well as their type (empirical or theoretical).
First, still in Section 3, each system will be presented in detail (Section 3). As we identified a strong tendency that new PCS are card-based (completely or at least a part of it), we thought it would be interesting to discuss individually the systems dividing them into 3 groups: card-based (3.1 to 3.8), partially card-based (3.9 to 3.11) and non-card-based (3.12 and 3.13).
Next, in Section 4, the PCS’ will be compared regarding all four variables of systems characteristics and the two of systems evolution, in order to draw some conclusions on their similarities and differences.
Inverse Base Stock (IBS)
Little explored in literature, Inverse Base Stock (IBS) was proposed by Masin, Herer and Dar-El (1999). Apart from its conceptual proposal, there is no other study about IBS in literature. Therefore, this system stops at a very early stage.
IBS is part of the self-regulated WIP (SWIP) approach, also proposed by Masin, Herer and Dar-El (1999), which unifies several PCS such as Kanban, CONWIP, Drum-Buffer-Rope (DBR), Base Stock, among others. The main feature of SWIP is to group a set of equipment into a subsystem that shares the same number of containers or cards. In CONWIP, for example, the entire system shares the same number of containers, while in Kanban each pair of adjacent workstations is a subsystem.
The name Inverse Base Stock is due to the visual representation of this system, which is the mirror image of Base Stock (Figure 4). IBS releases a job on the first station only if cards are available for processing that order at all stations in the system. After being processed in a station, the order releases the card of that station.
Table 2 - The 13 PCS’s classified according to the four variables of system’s characteristics dimension
System Acronyms Year Main Reference Primary Control
Variable Level of Centralization Complexity of material flow Card-based
Inverse Base Stock IBS 1999 Masin (1999) WIP Centralized Flow shop Yes
CONstant LOAD CONLOAD 1999 Rose (1999) WIP Centralized Flow shop No
Customize
Token-based system CTBS 2000
Gaury, Pierreval,
and Kleijnen (2000) WIP Both Flow shop Yes
Double Speed Single
Production Line DSSPL 2000
Stagno, Glardon, and
Pouly (2000) WIP Centralized Flow shop Partially
Decentralised Work in
Process DEWIP 2000
Lödding and
Wiendahl (2000) WIP Decentralized Job shop No
Behaviour Based
Control BBC 2001
Paternina-Arboleda
and Das (2001) WIP Both Flow shop Yes
Gated MaxWIP G-MaxWIP 2002 Grosfeld-Nir and
Magazine (2002) WIP Centralized Flow shop Yes
Parallel Pull Flow PPF 2004 Hunter et al. (2004) WIP Centralized Flow shop Yes Control of Balance by
Card Based Navigation
COBACABA
NA 2009 Land (2009) WIP Centralized Job shop Yes
Demand Driven Materials Requirement Planning
DDMRP 2011 Ptak and Smith
(2011) Throughput Centralized Job shop Partially Basestock
Kanban-Constant Work-in- Process
BK-CONWIP 2012 Onyeocha and
Geraghty (2012) WIP Both Flow shop Yes
Redutex - 2016 Serrato (2016) Throughput Centralized Flow shop Partially
Basestock-Constant
Work-in- Process B-CONWIP 2018
Hawari, Qasem, and
Figure 3- Production Control Systems evolution from 1999 to 2018
Axiomatic and empirical studies Only empirical studies
Only axiomatic studies IBS (1999) Masin, Herer and Dar-El (1999) Masin, Herer and Dar-El (2005) CONLOAD (1999) Rose (1999) Rose (2001) CTBS (2000) Gaury, Pierreval and Kleijnen (2000) Gonzáles-R et al. (2007) González-R and Framinan (2009) DSSPL (2000) Stagno, Glardon and Pouly (2000) Cheikhrouhou (2007) Cheikhrouhou, Hachen and Glardon (2009) Cheikhrouhou, Hachen and Glardon (2012) DEWIP (2000) Lödding and Wiendahl (2000) BBC (2001) Paternina-Arboleda and Das (2001) G-MaxWIP (2000) Grosfeld-Nir and Magazine (2002) Sepehri and Nahavandi (2007) PPF (2004) Hunter et al. (2004) Hunter (2006) Lasa, Vila and Uriarte (2009) COBACABANA (2009) Land (2009) Thürer, Land and Stevenson (2014) Thürer, Stevenson and Protzman (2016) Thürer, Stevenson and Protzman (2017) DDMRP (2011) Ptak and Smith (2011) Miclo et al. (2016) Ptak and Smith
(2016) Miclo et al. (2019) Redutex (2016) Serrato (2016) Hamja et al. (2017) Serrato (2018) BK-CONWIP (2012) Onyeocha and Geraghty (2012) Onyeocha, Khoury and Geraghtu (2015) Onyeocha et al. (2015) B-CONWIP (2018) Hawari, Qasem and Smadi (2018) Lödding, Yu and Wiendahl (2003) Thürer, Stevenson and
Protzman (2015) Onyeocha and Geraghty (2013)
Figure 4 - Base Stock and Inverse Base Stock
Source: Adapted from Masin, Herer and Dar-El (1999) and Masin, Herer and Dar-El (2005)
Customised token-based systems (CTBS)
Proposed by Gaury, Pierreval and Kleijnen (2000), the Customized token-based systems (CTBS) is a generalization of token-based systems, which generally use cards as token (GONZÁLES-R et al., 2007). According to Liberopoulos and Dallery (2000), this class of PCS is the easiest to implement and the most studied in the literature. As shown in Figure 5, CTBS considers all possible relationships between workstations. Specific systems, such as the CONWIP (loop k13 - between the first and last station), are CTBS special cases (GONZÁLES-R et al., 2007).
Figure 5 - Customised token-based systems
Source: Adapted from González-R et al. (2007).
In general, the selection of a PCS is based on an a priori approach, which means that a PCS is selected without considering the specific characteristics of the factory floor, such as
DEMAND
Base Stock
Inverse Base Stock
DEMAND DEMAND 1 2 3 k11 k22 k33 k12 k23 k13
processing times, set-up times, demand, machines workload, among others (GONZÁLEZ-R; FRAMINAN, 2009). CTBS, on the other hand, is based on a posteriori approach, which starts from a generic system that is shaped from the environmental knowledge. Therefore, CTBS consider all the space solutions in order to choose the best set of parameters (all possible loop structures and how many cards to keep in each of them).
To give an example, if CONWIP was chosen as the PCS, then only loop k13 will exist. Therefore, the task is to define how many cards to maintain in this loop. In an a posteriori approach, on the other hand, it is considered which loop structures should exist and how many cards to keep in each of them. González-R and Framinan (2009) accomplish this task using the cross-entropy method.
After its conceptual proposal, this system was further developed by Gonzales-R et al. (2007) and Gonzáles-R and Framinan (2009) which compared CTBS with other systems by simulation and proposed a method to develop the a posteriori approach using cross-entropy. However, no empirical study of this system was reported in literature.
Behaviour-Based Control (BBC)
The Behavior-Based Control (BBC) system was proposed by Paternina-Arboleda and Das (2001). BBC is based on the reinforcement learning concept, in which decision-makers learn optimal control policies by receiving rewards and punishments as a result of their actions. Therefore, decision-maker chooses actions to maximize their rewards over time (KAELBLING; LITTMAN; MOORE, 1996)
The system has three types of authorizations (Figure 6):
• CONWIP authorization: whenever a demand is met, the CONWIP card returns to the first stage, authorizing the production of a new item;
• Kanban authorization: at all except for the last stage, there are kanban cards to restrict the buffer between stages;
• Emergency authorization: whenever a demand is not met or a machine breaks, an emergency authorization card is released. This card authorizes the production of an additional unit and cannot be reused.
Using simulation, Paternina-Arboleda and Das (2001) showed that in a repetitive flow shop environment, BBC presents better performance than other systems, such as Kanban, CONWIP Base Stock, Extended Kanban Control System (EKCS) and two-boundary hybrid control. However, this system was also not further developed and lack empirical studies to prove it can be useful in practice.
Figure 6 - Behaviour-Based Control authorizations
Source: Adapted from Paternina-Arboleda and Das (2001).
Gated MaxWIP (G-MaxWIP)
Gated MaxWIP (G-MaxWIP) is a hybrid PCS proposed by Grosfeld-Nir and Magazine (2002) in which all production stages are pushed, except for the first, which is pulled. The main characteristics of this system is that the first production stage is used as gate (Figure 7). This gate controls the entrance of materials into the system based on the system WIP. If WIP is below a certain defined level, the gate stays opened and lets materials enter the system. When WIP reaches a pre-set maximum WIP level, the gate closes. Then, two strategies can be used to open the gate: as soon as the WIP reaches a certain level or after a certain time interval.
Figure 7 – Gated MaxWIP
Source: Elaborated by the author.
According to Grosfeld-Nir and Magazine (2002), G-MaxWIP combines two of the most desirable features of PCS. The first one (pull) is to control the system WIP by opening and closing the gate. Regarding this point, G-Max WIP and CONWIP work similarly. The second one (push) is to allow resources to work unrestricted, increasing utilization. This is true, unless when the gate is closed. In this moment, some stages can become idle due to the lack of material to be processed (SEPEHRI; NAHAVANDI, 2007).
Sepehri and Nahavandi (2007) compared G-MaxWIP with CONWIP and CWIPL (critical WIP loops) through simulation studies, however no other development of this system
DEMAND (satisfied) Legend: Kanban Emergency Authorizations ConWIP Lost Demand DEMAND GATE
was found in literature. Moreover, G-MaxWIP also lacks empirical studies.
Parallel Pull Flow (PPF)
Based on Lean principles, Parallel Pull Flow (PPF) was developed by Hunter et al. (2004) and it was not found any other reference to this system in literature, apart from Lasa, Vila and Uriarte (2009). Developed originally for furniture and wood components industry, PPF consists of a return-loop (rectangular or oval configuration), in which one side is used for kitting and staging carts and the other for final assembly.
When an item is assembled, the container returns empty to the purchased components area, where the necessary components for the assembly of the next final product are collected (Figure 8). The container also collects the required semi-finished items produced in subassembly lines. With all the necessary components, the container enters the assembly line and a sequence of activities is performed. Once the item is assembled, the final product is delivered and the container returns empty to the component area, collecting the necessary components for the next order (HUNTER; 2006).
Figure 8 - Parallel Pull Flow
Source: Adapted from Hunter (2004).
This system, little explored in practice and literature, is argued to be useful for environments where component availability is critical. The coordination between the subassembly lines and the assembly line can be performed by another PCS. Hunter et al. (2004)
Empt y C ontaine rs Kitt ed Components Final Assembly Subassembly Kitting Demand 1 2 3 4 5 6 7 8
suggests the use of Kanban systems.
Control of Balance by Card Base Navigation (COBACABANA)
COBACABANA, an acronym for Control of Balance by Card Based Navigation, was proposed by Land (2009) and refined by Thürer, Land and Stevenson (2014). Unlike other card-based systems, COBACABANA uses the Workload Control approach, releasing orders based on the workload of critical stations (THÜRER; STEVENSON; PROTZMAN, 2017). Therefore, COBACANABA creates a card loop between the central planner and critical workstations (Figure 9).
By controlling the workload at stations, COBACABANA also focuses on controlling the throughput times of each station (LAND, 2009). COBACABANA uses a pair of cards:
• Release card: this card stays with the central planner and is used to calculate the workload in the shop-floor;
• Operation card: this card goes to the shop floor with the released order and return to the central planner after an operation is complete.
Figure 9 - COBACABANA card loop
Source: Adapted from Thürer, Land and Stevenson (2014).
COBACABANA uses a centralized release orders function called pre-shop pool. In this pool, orders are sorted according to their due date (THÜRER; LAND; STEVENSON, 2014). Before releasing an order on the shop floor, the planner evaluate whether this order will not
1 2 3 Central Planner Operation Cards Planning Board Release Cards
exceed the capacity limits set for critical stations. If the order violates these limits, the planner considers releasing the second order and so on, until all the orders in the pool were considered. To assess if there is available capacity at the stations, the planner compares the current workload with the workload limits on the planning board (Figure 10). For each operation card release, a release card (of the same workload) is placed on the board. Each time an operation card returns to the planner, a release card with the same workload is removed from the board.
According to Thürer, Stevenson and Protzman (2017), orders in COBACABANA can be released periodically (Original COBACABANA) or continuously (Continuous COBACABANA). In the first case, orders are released at fixed time intervals or also releases an order without load considerations whenever the first station in the routing of the order is starving. In the second case, release decisions are taken whenever an operation is completed or a new order arrives at the pre-shop pool.
Figure 10 - COBACABANA planning board
Source: Adapted from Thürer, Land, and Stevenson (2014).
Regarding the environment, COBACABANA was originally proposed for high-variety job shop contexts, but studies show a good system performance even in pure flow shop (THÜRER; STEVENSON; PROTZMAN, 2015). However, to the best of our knowledge, there is no empirical study of this system was reported in literature.
Released Cards Allowance for Release Work Center 3 Work Center 2 Work Center 1 Workload Norm 20% 40% 60% 80% 100%
Basestock Kanban-CONWIP (BK-CONWIP)
BK-CONWIP was developed from HK-CONWIP (Hybrid Kanban-CONWIP). The HK-CONWIP was proposed by Bonvik, Couch and Gershwin (1997), with the aim of offering a system that would respond to an environment with a greater variety of products, controlling the total inventory of the system (CONWIP cards), but also the stock of each stage, except the last one (Kanban cards). Therefore, two production authorization cards are used. Several studies show that HK-CONWIP performs better Kanban and CONWIP, such as Geraghty and Heavey (2004) and Wang, Cao and Kong (2009). HK-CONWIP was originally developed for a single product and several studies have assumed the possibility of replicating it for various products (ONYEOCHA et al., 2015).
In a multi-product PCS, two different authorization policies can be used: Shared Kanban Distribution Policy (S-KDP) or Dedicated Kanban Distribution Policy (D-KDP) (BAYANT; BYZACOTT; DALLERY, 2002). While in D-KDP each card is specific to authorize the production of a single product, in S-KDP a card can be shared by a set of items (in this policy, the specific item to be produced is selected according to demand and materials availability). This makes S-KDP more flexible to variations in demand than D-KDP.
As shown by Bayant, Buzacott and Dallery (2002), Onyeocha and Geraghty (2012) and Olaitan and Geraghty (2013), several pull systems such as Kanban, CONWIP, Base Stock and HK-CONWIP have a bad performance when operating S-KDP in a multi-product environment, due to the method they adopt to transmit demand variations to the system. In this context, Onyeocha and Geraghty (2012) proposed the Basestock Kanban-CONWIP (BK-CONWIP) as an alternative to HK-CONWIP, in order to allow this system to work with the S-KDP policy. As Onyeocha, Khoury and Geraghtu (2015) state, BK-CONWIP is suitable for environments with high production mix flexibility and can operate with both the S-KDP strategy and the D-KDP strategy.
In BK-CONWIP, demand information is globally transmitted to all production stages (ONYEOCHA et al., 2015). The system has three control parameters (CONWIP cards, Kanban cards and Stock levels). As well as in the HK-CONWIP, CONWIP authorization cards are used to control the stock of the whole system and Kanban authorization cards to control the inventory on each stage. The Base Stock level in finished products is used to control the overall flow of demand information into the system.
When an order enters the system, demand information is sent to all production stages and special information to the last stage, so that it releases a CONWIP card to satisfy the demand. If raw materials and capacity are available, production starts simultaneously at all
stages. If not, production starts, but is interrupted at the stage with capacity restriction or lack raw materials. As soon as the final product arrives in stock, the CONWIP card returns to the buffer (Figure 11).
Figure 11 - BK-CONWIP
Source: Elaborated by the author.
Apart from Onyeocha and Geraghty’s articles, other references to BK-CONWIP are restricted to conferences. However, this system presents an interesting initial evolution, which includes a simulation study (ONYEOCHA et al., 2015). However, this system also lacks empirical studies in order to understand how it will react to real production environments.
Basestock-CONWIP (B-CONWIP)
The Basestock-Constant Work-In-Process (B-CONWIP) was proposed by Hawari, Qasem and Smadi (2018) from BK-CONWIP. Apart from its proposal article using simulation, no other reference to B-CONWIP was found in literature. The objective of this system is to minimize WIP and to achieve specified service levels. Like BK-CONWIP, B-CONWIP can operate with S-KDP and D-KDP policy, and has two control parameters:
• Base Stock levels: Minimum inventory level at each stage so that it meets all unanticipated demand;
• CONWIP authorization card: limits WIP throughout the system.
The main difference between BK-CONWIP and B-CONWIP is that B-CONWIP does not use kanban cards between the stations (Figure 12). A balancing algorithm to control the stock of each productive stage is uses instead. According to Hawari, Qasem and Smadi (2018), the control of WIP levels of both systems is similar, with the advantage of B-CONWIP being simpler, especially in environments with many productive stages and many products.
DEMAND Legend:
Kanban
Base Stock ConWIP
Figure 12 - B-CONWIP
Source: Elaborated by the author.
The authors also claim that when demands increase, B-CONWIP is more appropriate if the priority is the service level, while BK-CONWIP is the best option if WIP control is the most important variable.
Double Speed Single Production Line (DSSPL)
The Double Speed Single Production Line (DSSPL) is a hybrid PCS proposed by Stagno, Glardon and Pouly (2000). It was developed for industries with many distinct products (and small variety among them) and a wide variation in demand. Its main distinction from others PCS is its selectivity in allocating products to resources. Items are segregated into two groups:
• A-products: small number of products, with high production volume and fairly regular demand;
• B-products: large number of products, sold in small quantities and with irregular demand.
In DSSPL, items A are produced quickly through a pull system and items B are controlled by a push system (Figure 13). Through this segregation, it is possible to reduce lead time and stock levels of items A without significantly affecting items B. However, since items A correspond to a high volume, this change has a significant impact on the overall system result. An application of DSSPL is presented by Cheikhrouhou, Hachen and Glardon (2009).
Stagno, Glardon and Pouly (2000) also mention that the classification criteria can be the type of customer, with A clients being the most important ones. Therefore, a product is A when produced for some clients and B when produced for others. Other classification criteria are also possible. However, an assumption of DSSPL is that demand for A-products is sufficiently stable, otherwise a pull system could not be successfully implemented.
DEMAND Legend:
Base Stock ConWIP
Figure 13 - DSSPL
Source: Adapted from Cheikhrouhou (2007).
This system was also study by Cheikhrouhou, which compared DSSPL through simulation with other PCS (CHEIKHROUHOU, 2007; CHEIKHROUHOU; HACHEN; GLARDON, 2012). However, no empirical study was reported in literature.
Demand Driven Materials Requirement Planning (DDMRP)
Demand Driven Materials Requirement Planning (DDMRP) is a hybrid PCS proposed by Ptak and Smith (2011). This system aims to combine the best practices of MRP II, Lean Manufacturing and Theory of Constraints (MICLO et al., 2019). According to Miclo et al. (2016), DDMRP has been developed since 2000 and has already been implemented in some United States companies. In literature, empirical and simulation articles of DDMRP can be founded.
DDMRP is based on four basic principles (PTAK; SMITH, 2016):
• Decoupled Lead Time: Some pre-defined Bill of Materials components are kept in stock (in Figure 14, items D and F are kept in stock);
• Decoupled Explosion: For components held in stock, the requirements are not generated by the traditional MRP explosion, but by the ASE;
• Available Stock Equation (ASE): calculated daily, projects future stock based on actual demand (not forecast) and orders in production. The ASE is compared
DEMAND
Legend:
Demand A items (pull system) Demand B items (push system)
to 3 buffer levels: red (safety stock), yellow (every time the ASE reaches the yellow zone, a new order is released for the stock to reach the top of the green zone) and green (replenishment size);
• Relative priority: color of the card according to the zone in which the item is located. The orders also show a percentage of the stock projected by the ASE vs maximum stock of the item (top of the green zone).
Figure 14 - DDMRP
Source: Elaborated by the author.
According to Ptak and Smith (2011), the implementation of DDMRP occurs in 5 steps. They are divided into: modeling the environment (Steps 1, 2 and 3), Plan (4) and Execute (5). The stages are:
(1) Strategic Inventory Position: to evaluate, from a financial point of view, if an item of the Bill of Materials should or not be maintain in stock. The main function of the buffer is to absorb variability. Therefore, unlike a normal MRP, in DDMRP unbuffered items are pushed, but buffered items are pulled, replenishing inventory;
(2) Buffer Profiles and Levels: to size green, yellow and red zones based on the following equations:
GreenZone = Max (YellowZone x Lead Time Factor; LotSize) (1)
Legend: MRP Explosion ASE Equation A B C F D E
YellowZone = ADU x ASRLT x PAF (2) RedZone = YellowZone x LTFactor x (1+Variability Factor) (3) Which:
• ADU (Average Daily Usage): daily average demand, estimated by demand forecast;
• ASRLT (Actively Synchronized Replenishment Lead Time): the longest unprotected sequence, considering the sum of lead time of the bill of material, of a buffered item;
• PAF (Plan Adjustment Factors): used to raise or lower the ADU, allowing to smooth seasonality. It should be defined based on the master plan capacity analysis.
(3) Dynamic Adjustments: to adjust the zones with changes in sales forecast; (4) Demand Driven Planning: to create production and purchase orders; (5) Visible and Collaborative Execution: to control the orders generated.
REDUTEX
REDUTEX is a hybrid system developed by Serrato (2016) which aims to reduce customer lead time. It consists of 8 steps and is based on Lean (steps 4, 5 and 8) and Theory of Constraint (steps 1, 2, 3, 4 and 6) principles. Step 7 is particular of REDUTEX. The focus of the system is small and medium enterprises (SME) in which low technological knowledge is used (TENG; JARAMILLO, 2006); for example the textile industries in Central Mexico.
The sequence of REDUTEX steps are:
(1) Identify the Restrictive Resource Capacity (RRC): to identify the resource that need to work with 100% of the daily capacity to meet demand;
(2) RRC optimization: to optimize the set ups in the RRC to increase the total production of the system;
(3) Synchronize rhythm with the RRC: all other resources must work at the same pace of the RRC. To do this, is necessary to adjust work shifts. Half shifts can be used (half of the time on one equipment, half on another);
(4) Create a smooth and continuous flow throughout the process: material must flow gradually into the system and there should be no accumulation of inventory between departments. To do so, it is essential to define an appropriate transference batch between processes (Figure 15);
concept that the more, the better, because it ensures RRC to work at full capacity. All resources before the supermarket (RRC included) are pushed and the ones after it are pulled;
(6) Create a time buffer: The time buffer is usually located before an assembly department and is a protection against fluctuations and delays in previous processes. In the example of Serrato (2016), it is located together with the supermarket. The time buffer purpose is to ensure that the components necessary to next day program are available. If not, the supervisor must verify at which point of the process they are and which actions are necessary to make them available at the assembly time;
(7) Control production through automated dual card system control: the system uses cards to identify and track products in the factory. The card follows the flow of the product throughout the entire factory and is transferred to the next department when the entire lot has been processed in the previous one. Regarding the card design, the right side contains barcodes and the left side information about each department. The card specifies product’s type, size, color, department, production lot, transfer lot and operator;
(8) Visual quality control: in a board for each department, the results of batch inspection are visually displayed. A green point indicates a batch that meet specifications, yellow one within specifications limit and red a batch that does not meet specifications. Serrato (2016) suggests organizing the board in this way: columns (types of defects evaluated) and lines (batches evaluated).
Figure 15 – Redutex
Source: Elaborated by the author.
Even though it is a recent PCS, other references to Redutex include Hamja et al. (2017) and Serrato (2018), however none of these articles contributes to further developed Redutex. Therefore, this system first lacks simulation studies comparing it to other PCS. Secondly, it also lacks empirical studies, as Serrato (2016) is a unique example.
DEMAND RR C Time Buffer Assembly RRP Supermarket
Constant Load (CONLOAD)
CONstant LOAD (CONLOAD) is a PCS developed by Rose (1999) to overcome the difficulty of other systems, such as CONWIP, to deal with changes in product mix, in the transition period when one item is discontinued and another is introduced (ROSE, 2001). Such changes are very frequent in the semiconductor industry, where there are a large number of products with a very short life cycle due to technological changes.
CONLOAD was developed merging concepts of CONWIP and Workload Control (ROSE, 1999). Instead of controlling the WIP (like CONWIP), CONLOAD controls the bottleneck load. The bottleneck load is equal to the processing times in the bottleneck of all orders that already have been released but have not yet been processed in the bottleneck. Therefore, a job enters the system only if its processing time in the bottleneck plus the processing time in the bottleneck of all orders already released do not exceed a predefined workload (Figure 16).
A constraint of CONLOAD is the necessity to know products’ cycle times with high accuracy (ROSE, 2001). If this cycle time is overestimated, the bottleneck will become idle. If it is underestimated, the bottleneck will be overloaded and there will be accumulation of orders in front of this resource.
Figure 16 – CONLOAD
Source: Elaborated by the author.
Rose (1999) compared CONLOAD to CONWIP and Workload Regulation and found that CONLOAD is more efficient in maintaining the utilization level of the bottleneck while WIP evolves more smoothly over time.
DEMAND Legend:
Bottleneck
Processing times in the bottleneck of orders already released Central
control of order release
CONLOAD was little study in literature and, as many other systems of this review, lacks empirical studies.
Decentralized Work in Process (DEWIP)
The Decentralized Work in Process (DEWIP) was developed by Lödding and Wiendahl (2000) for job shop environments with the aim of offering smaller and more reliable lead times (LÖDDING; YU; WIENDAHL, 2003). Although there are many simulation studies about this PCS, no empirical study was found in literature.
The motivation to develop DEWIP arose from the fact that, although there are several decentralized systems whose primary control variable is WIP (Kanban and POLCA, for example), none of them is suitable for environments with complex flow of materials.
In DEWIP, all orders are programmed by a Central PCP, which sets production priorities. However, the actual moment when each operation starts is controlled by WIP as follows:
• The operator of a work center A checks the first order that needs to be produced and asks for authorization for the next work center (go-ahead request) (Figure 17);
Figure 17 – DEWIP
Source: Elaborated by the author.
• The downstream work center operator (B) verifies the workload of its own center (direct WIP) as well as the production authorizations already provided to upstream centers (indirect WIP). If releasing the new order, the total WIP (direct + indirect) exceeds a pre-set threshold, authorization is not provided;
• If authorization is provided, center A starts production and reduces its WIP,
A B C D E F
providing authorizations for the upstream work centers;
• If authorization is not provided, the operator of center A searches for the next order that is not processed in center B (which is possible because the environment is a job shop) and requests authorization for another center (C). This evaluation is performed in all work centers, establishing control loops between them. The only exception is critical resources, for which authorizations are always provided.
2.4 Comparison of New Production Control Systems (PCS)
After presenting the 13 PCS’s and classifying them according to the six variables defined in section 2.2 (Table 2 and Figure 2), the main findings will be discussed. Regarding the systems evolution dimension, first it can be observed that each system was developed by only one or two groups of authors. These can be one of the reasons why these systems is still little known in practice.
Secondly, it is important to highlight that while some systems were developed almost 20 years ago others are much more recent. Therefore, while B-CONWIP still have a high probability to thrive, chances for IBS are much slower. In our analysis, we could not find any prediction to the success of a PCS’s, however it may be due to the systems characteristics itself, to the journal it was published, to the group of authors that proposed the system, among other possibilities.
Thirdly, we observed that almost all PCS have been developed only in theory, specifically by mathematical simulations. PPF, DDMRP and Redutex are the only counterexamples, that is, systems that have been developed from practice. This can be explained by the advances in computing, which made simulation faster and able to work with more data, and, therefore, closer to reality. However, this scenario also led to an unwanted effect, that is, many PCS do not have empirical studies showing their application in practice. Therefore, theory and practice of PCS may be taken different paths.
Fourthly, we only find theoretical and empirical studies about one of the 13 PCS’s (DDMRP). This is a problem even for PPF and Redutex, as theoretical studies, such as computer simulation, could help to optimize systems parameters, increasing the chances of an empirical successful implementation of these PCS’s. Moreover, it reinforces the idea of theory and practice of PCS following different paths.
Regarding the system’s characteristics dimension, first we noticed predominance of systems (7 out of 13) which present WIP as primary control variable, are designed for flow shop environments and are card-based. In our understanding this is due to Lean influence, as
